Automatic Transfer Function Design for Medical Direct Volume Rendering via Clustering-Based Ray Analysis

Abstract

Transfer Function (TF) design is a central topic in medical direct volume rendering (DVR). TF design allows for interactive identification of features of interest (FOIs) within a medical image volume and their visual emphasis by assigning appropriate optical parameters (opacity and color) to them. Conventional TF design, however, is not intuitive and usually a 'trial-and-error' process for most users. In this work, an automatic TF design scheme is proposed which consists of two-steps. First, I introduce a new clustering-based ray analysis (CRA) to automatically identify FOls along a viewing ray defined by users. Here, the proposed CRA approach uses regional and contextual information around rays to improve the identification capability. Second, the proposed CRA approach automatically generates a TF to emphasize identified FOls by adopting a visibility-driven TF parameter optimization algorithm. Experiments show the effectiveness of the proposed CRA approach by demonstrating its advantages over the existing ray analysis approach relying on local intensity profiles of a ray. I evaluate a number of medical image volume datasets to show the utility of the proposed CRA approach for automatic TF design.